Accession Number : ADA614562


Title :   981: Evaluation of Burn Sepsis Automated Alerts in an Intensive Care Unit


Descriptive Note : Abstract


Corporate Author : ARMY INST OF SURGICAL RESEARCH FORT SAM HOUSTON TX


Personal Author(s) : Mann-Salinas, Elizabeth ; Caldwell, Nicole ; Serio-Melvin, Maria ; Luellen, David ; Chung, Kevin ; Cancio, Leopoldo ; Salinas, Jose


Full Text : http://www.dtic.mil/get-tr-doc/pdf?AD=ADA614562


Report Date : 01 Dec 2014


Pagination or Media Count : 2


Abstract : Learning Objectives: In spring 2013 the burn unit initiated Sepsis Alert software. Continuous electronic medical record (EMR) screening used novel predictors of burn sepsis (Burn6 Criteria): heart rate (HR) 130 bpm, mean arterial pressure (MAP) 60 mmHg, base deficit (BD) -6 mEq/L, temperature (T) 36 C, use of vasoactive medications (VM), or glucose 150 mg/dl. We hypothesize sepsis alerts will be both accurate and meaningful to the provider to avoid alert fatigue prior to expanding the number of clinical staff receiving the alerts. Methods: Prospective observational study of alert performance was conducted. An automated email was sent to identified personnel. Simple rules included: /= 2 of the parameters are out of range; patient /= 48 hours post admission; /= 12 hours post-operative; and admitted to the intensive care unit. EMR was reviewed for clinical applicability for each alert. Descriptive and non-parametric statistics were used to describe variables when alert triggered vs not triggered. Results: From 8/2013 7/2013 149 alerts were generated on 47 patients (3.7 3.6; range1-16 alerts/patient) with 24 16% total burn surface area. Frequencies of variables in alerts, with value mean SD for triggered vs not triggered respectively, were: glucose 67%, 181 32 vs 117 20; MAP 67%, 56 4 vs 81 14; VM 48%, 1.4 0.6 vs 0 0; HR 27%, 139 14 vs 97 18; T 15%, 35.6 0.4 vs 38 0.9; and BD 4%, -8.2 3 vs 3.7 4 (all p values within groups 0.0001). Each alert contained 2.1 0.31 (range 2 4) variables. Patients were receiving antibiotics and antifungals concurrent with 50/71 and 21/70 alerts, respectively. Cultures were obtained after 7/65 alerts, and prior to 27/65 alerts. Lactate was /= 2 mEq in 15/55 instances. Conclusions: BD and T appear to be poorly correlated with burn sepsis alerts. Most patients were receiving antimicrobial therapy when alert was triggered; a small number of alerts preceded culture acquisition.


Descriptors :   *BURNS(INJURIES) , *SEPSIS , AUTOMATIC TRACKING , INTENSIVE CARE UNITS , MONITORING


Subject Categories : Medicine and Medical Research
      Stress Physiology


Distribution Statement : APPROVED FOR PUBLIC RELEASE